Localization of Caspr2 in Myelinated Nerves Depends on Axon–Glia Interactions and the Generation of Barriers Along the Axon

The Journal of Neuroscience, October 1, 2001, 21(19):7568–7575

Localization of Caspr2 in Myelinated Nerves Depends on Axon–Glia Interactions and the Generation of Barriers along the Axon

Sebastian Poliak,1 Leora Gollan,1 Daniela Salomon,1 Erik O. Berglund,2 Reiko Ohara,3 Barbara Ranscht,2 and Elior Peles1 1Department of Molecular Cell Biology, The Weizmann Institute of Science, Rehovot 76100, Israel, 2The Burnham Institute, Neurobiology Program, La Jolla, California 92037, and 3Department of Human Gene Research at Kazusa DNA Research Institute, Chiba 292-0812, Japan

Cell recognition proteins of the contactin-associated protein before relocating to the adjacent juxtaparanodal region. This (Caspr) family demarcate distinct domains along myelinated transition was not observed in CGT mice, where Caspr2 and axons. Caspr is present at the paranodal junction formed be- Kv1.2 remained paranodal. Double labeling for Caspr and tween the axon and myelinating glial cells, whereas Caspr2 is Caspr2 demonstrated that these two related proteins occupied localized and associates with K ϩ channels at the adjacent mutually excluding domains along the axon and revealed the juxtaparanodal region. Here we investigated the distribution of presence of both paranodal and internodal barrier-like struc- Caspr2 during development of peripheral nerves of normal and tures that are delineated by Caspr. Finally, we found that the galactolipids-deﬁcient [ceramide galactosyl transferase disruption of axon–glia contact in CGTϪ/Ϫ nerves also affects (CGT)Ϫ/Ϫ] mice. This mutant exhibits paranodal abnormalities, the localization of the cytoskeleton-associated protein 4.1B lacking all putative adhesion components of this junction, in- along the axon. Altogether, our results reveal a sequential ap- cluding Caspr, contactin, and neurofascin 155. In sciatic nerves pearance of members of the Caspr family at different domains of this mutant, Caspr2 was not found at the juxtaparanodal along myelinated axons and suggest that the localization of region but was concentrated instead at the paranodes with Caspr2 may be controlled by the generation of Caspr- Kv1.2. Similar distribution of Caspr2 was found in the PNS of containing barriers along the axon. contactin knock-out mice, which also lack Caspr in their paranodes. During development of wild-type peripheral nerves, Key words: Caspr2; myelin; node of Ranvier; axo-glial Caspr2 and Kv1.2 were initially detected at the paranodes junction; Schwann cells; K ϩ channels

Saltatory conduction depends on the generation of specialized point of contact that is formed between the axolemma and the domains along myelinated axons in which different sets of ion paranodal loops of the myelinating cells (Rosenbluth, 1995). channels are present (Arroyo et al., 1999; Peles and Salzer, 2000). The axonal membrane at the paranodes contains a complex of At the node of Ranvier, the axonal membrane contains high cell adhesion molecules that include Caspr (also known as ϩ concentrations of voltage-gated Na channels that are responsi- paranodin) (Einheber et al., 1997; Menegoz et al., 1997; Peles ble for the regeneration of the action potential (Waxman and et al., 1997) and the glycosylphosphatidyl inositol-linked cell Ritchie, 1993). The juxtaparanodal region is located beneath the adhesion molecule contactin (Rios et al., 2000; Boyle et al., compact myelin at both sides of each internodal interval and 2001). A glial component of the paranodal junction was re- ϩ contains delayed rectifier K channels of the Shaker family, cently identified as neurofascin 155 (NF155), a spliced isoform Kv1.1, Kv1.2, and their Kv␤2 subunit (Wang et al., 1993; Rhodes of the cell adhesion molecule neurofascin that is specifically et al., 1997; Vabnick and Shrager, 1998; Rasband and Shrager, found at the glial loops (Tait et al., 2000). A schematic diagram 2000). These channels colocalize and create a complex with summarizing the proteins found in the various subdomains at Caspr2, the second member of a growing family of putative cell the nodal region is presented in Figure 9. recognition molecules that demarcate distinct domains along my- ϩ The generation and maintenance of specialized domains in elinated fibers (Poliak et al., 1999). The juxtaparanodal K chan- ϩ myelinated fibers are controlled by glial contact and intrinsic nels and Caspr2 are separated from the Na channels at the node determinants within the axon (Peles and Salzer, 2000). Although of Ranvier by the paranodal junction, a specialized septate-like the exact mechanisms involved are not completely understood, such organization may require the presence of barriers along the Received May 30, 2001; revised July 12, 2001; accepted July 18, 2001. axon that restrict the diffusion of axonal proteins (Winckler et al., This work was supported by grants from the National Multiple Sclerosis Society 1999). Such barriers may be present at the paranodal junction. (Grant RG-3102), the United States–Israel Science Foundation (BSF), Jerusalem, Israel, and the Minerva Foundation to E.P. and from the National Institutes of Although structurally distinct, the paranodal junctions were sug- Health (NS38397) and the March of Dimes Birth Defects Foundation (FY00-240) to gested to function similarly to tight junctions by providing a fence B.R. E.P. is an Incumbent of the Madeleine Haas Russell Career Development Chair. We thank Brian Popko for his generous support in making the CGTϪ/Ϫ mice that restricts the movement of membrane proteins, thereby con- available to us and Jim Trimmer for his kind supply of antibodies to ion channels. fining them to specific domains along the axolemma (Rosenbluth, Correspondence should be addressed to Dr. Elior Peles, Department of Molecular 1995). The importance of axo-glial contact for the proper gener- Cell Biology, The Weizmann Institute of Science, Rehovot 76100, Israel. E-mail: [email protected]. ation of axonal domains is further supported by the analysis of Copyright © 2001 Society for Neuroscience 0270-6474/01/217568-08$15.00/0 mutant mice with defective paranodal junctions. These include Poliak et al. • Caspr-Dependent Localization of Caspr2 J. Neurosci., October 1, 2001, 21(19):7568–7575 7569 mice deficient for contactin (Berglund et al., 1999; Boyle et al., RESULTS 2001) and galactolipids-deficient mice (Dupree et al., 1999). The Paranodal junction components in CGT and latter lack the enzyme UDP-galactose/ceramide galactosyl trans- contactin mutants ferase (CGT) and are incapable of synthesizing the two major Previous work showed that Caspr is either absent or only weakly myelin galactolipids galactocerebroside (GalC) and sulfatide (Co- present in the paranodes of galactolipids mutant mice (Dupree et etzee et al., 1996; Bosio et al., 1998). Both of these paranodal al., 1999). It was suggested that the paranodal defects observed in mutants lack the transverse bands that are thought to represent this mutant result from inappropriate targeting of a glial adhesion the septate-like junctions in peripheral myelinated nerves and component to the paranodal loops (Dupree et al., 1999; Popko, exhibit abnormal localization of Caspr and Kv1.1 along their 2000). To further characterize this mutant, we examined the axons (Bosio et al., 1998; Dupree et al., 1998, 1999; Boyle et al., distribution of contactin and NF155, two adhesion molecules that 2001). To further understand the molecular mechanisms involved are found at the neuronal and glial sides, respectively, of the in the generation and maintenance of distinct functional domains paranodal junction (Rios et al., 2000; Tait et al., 2000). Although along myelinated axons, we have investigated the distribution of in wild-type animals NF155 and contactin show a similar distri- the Caspr2/Kv1.2 complex during development of peripheral bution to Caspr at the paranodes and the mesaxon, in CGTϪ/Ϫ nerves in wild-type and paranodal mutant mice. Our findings mice these adhesion molecules were hardly detected in the para- suggest that the localization of these proteins depends on the nodal region (Fig. 1). As observed in wild-type nerves (Tait et al., presence of axonal barrier-like structures that are outlined by 2000), NF155 was still detected in the Schmidt–Lanterman inci- Caspr at the paranodal junction and along the internodal sures (data not shown), indicating that it was specifically mislo- membrane. calized at the paranodes and the mesaxon. Similarly, NF155 staining was markedly reduced in the paranodes of contactin MATERIALS AND METHODS mutant mice, which also exhibit disruption of the paranodal septate-like junction (Fig. 1) (Boyle et al., 2001). In both para- Animals and antibodies. Generation and genotyping of CGT and contactin null animals have been described previously (Coetzee et al., 1996; nodal mutants, some residual NF155 staining could be detected Berglund et al., 1999). Wild-type control animals were derived from the occasionally in the paranodal loops, but it was clearly distinct same litters. Antibodies to Casprs include polyclonal antibodies P6061 from the wild-type nerve and did not appear to be concentrated and P4344 directed to the intracellular region of Caspr or Caspr2, at the paranodal junction. These observations indicate that local- respectively, as well as monoclonal antibody M275, which is directed to ization of the different adhesion components at the paranodal the extracellular domain of Caspr (Peles et al., 1997; Poliak et al., 1999). ϩ Monoclonal antibodies against Na channels and Kv1.2 (Bekele-Arcuri junction is interdependent, suggesting that galactolipids may af- et al., 1996; Rasband et al., 1999), as well as polyclonal antibodies against fect the local differentiation of myelinated axons by regulating the contactin (Berglund et al., 1999), were described previously. NF155- proper localization of NF155 to the glial loops. specific antibodies were generated by immunizing rabbits with a mixture of two peptides, IRVLNSTAISLQWNRVYPD and RQQASFPG- Distribution of Caspr2 in CGT knock-out mice ϩ DRPRGVVGRLF (amino acid positions 822–840 and 869–887, respec- The close association between Caspr2 and K channels raises the tively), derived from the third FNIII repeat of rat neurofascin (Davis et question of whether these proteins are still colocalized in myelin- al., 1996). ated nerves of CGTϪ/Ϫ and contactin null mice, which display Immunoprecipitation and immunoblot analysis. Coimmunoprecipita- ϩ tion experiments were done using brain membranes prepared from abnormal localization of K channels along their axons (Dupree wild-type or CGTϪ/Ϫ animals. Brains from four mice were pooled and et al., 1999; Boyle et al., 2001). In the sciatic nerves from these homogenized in a buffer containing 20 mM HEPES, pH 7.4, 0.32 M mutants, Caspr2 was found adjacent to the nodes of Ranvier, ␮ ϩ sucrose, 1 mM EGTA, 1.5 mM MgSO4, 10 g/ml aprotinin and leupeptin, which were labeled with a Na channel antibody (Fig. 2). Similar and1mM phenylmethyl-sulfonyl fluoride (PMSF). Nuclei and heavy cell ϩ debris were removed by low-speed centrifugation (1000 ϫ g) for 10 min, paranodal localization was found for K channels subunits Kv1.2 and the supernatant was subjected to high-speed centrifugation (Fig. 1) and Kv1.1 (data not shown) (Dupree et al., 1999; Boyle et ϩ (40,000 ϫ g) for 60 min. Membrane solubilization, immunoprecipitation, al., 2001). Although in wild-type animals Caspr2 and K chan- and immunoblots were conducted as described previously (Poliak et al., nels were present at 93% of the juxtaparanodes, they were local- 1999). ized at the paranodes in 94 and 96% of the sites in CGT and Immunofluorescence. Sciatic nerves were dissected and immersed in Zamboni’s fixative for 10 min. In some cases, nerves were first fixed in contactin mutant nerves, respectively. In the remaining 4–6% of 4% paraformaldehyde and teased before immersion in Zamboni’s fixa- the cases, these molecules were not at all or only weakly detected tive. After desheathing, nerves were teased on gelatin-coated slides, along the axon. Double labeling for Caspr2 and Kv1.2 demon- dried for 2 hr, and frozen at Ϫ20°C. Preparation and staining of optic strated that these proteins were colocalized at all of the sites ϩ nerve sections have been described previously (Poliak et al., 1999). The examined. In CGT and contactin mutant mice, Caspr2 and K preparations were post-fixed/permeabilized in methanol at Ϫ20°C for 20 min or permeabilized in PBS/0.5% Triton X-100 for 10 min for optimal channels did not occupy the entire paranodal length but were staining of 4.1B. Slides were washed and blocked for 30 min with PBS more concentrated near the nodes. This labeling pattern was containing 10% normal goat serum, 0.1% Triton X-100, and 1% glycine. clearly distinct from the distribution of Caspr in wild-type ani- The samples were incubated overnight at 4°C with the different primary mals. Altogether, these experiments demonstrate that Caspr2 is ϩ antibodies described above diluted in blocking solution. Slides were aberrantly relocated with K channels to the paranodes in two washed three times for 5 min in PBS and incubated for 1 hr in secondary antibodies: anti mouse-Cy3 or biotinylated anti-rabbit (Jackson Immu- different mutant mice that exhibit paranodal junction defects. noResearch, West Grove, PA). The latter was washed in PBS and reacted Thus, the presence of the Caspr/contactin complex in the paran- ϩ with streptavidin–Alexa-488 (Molecular Probes, Eugene, OR) for 45 min odes may restrict the movement of Caspr2/K channels toward and was further processed as described previously (Poliak et al., 1999). the nodes. Immunofluorescence was viewed and analyzed using a Deltavision wide- We have previously described that, in the CNS, Caspr2 colo- field deconvolution microscope (Applied Precision), using a Bio-Rad ␤ confocal microscope (Bio-Rad, Richmond, CA) or a Zeiss Axioplan calizes and forms a complex with Kv1.2, Kv1.1, and their Kv 2 microscope equipped with SPOT-II (Diagnostic Instruments) cooled subunits (Poliak et al., 1999). Myelinated fibers in the CNS of CCD camera. CGTϪ/Ϫ mice exhibit more severe morphological alterations as 7570 J. Neurosci., October 1, 2001, 21(19):7568–7575 Poliak et al. • Caspr-Dependent Localization of Caspr2

Figure 1. Aberrant localization of paranodal junction proteins in peripheral nerves of CGT and contactin knock-out mice. Teased sciatic nerves were prepared from either 25-d-old wild-type (WT) or CGT (CGTϪ/Ϫ) mice or from 17-d-old contactin knock-out (ConϪ/Ϫ) animals. The prep- ϩ arations were labeled with a mouse monoclonal antibody to Na channel (NaCh,inred) and an antibody to neurofascin 155 (NF155), Caspr, contactin, or Caspr2 (all in green), as indicated at the top of each panel. Scale bar, 10 ␮m. compared with the PNS (Dupree et al., 1998; Popko, 2000). Analyses of optic nerves of wild-type and CGTϪ/Ϫ mice revealed that, in contrast to the PNS, Caspr2 was found only infrequently at the juxtaparanodes, and in most ﬁbers was either not detected or diffused along the axon (Fig. 3A). A similar distribution was revealed using an antibody to Kv1.2. To determine whether ϩ ϩ Caspr2 and K channels associate in CGTϪ/Ϫ mice, brain mem- Figure 3. Caspr2/K channels complex is mislocalized in the CNS of CGT knock-out mice. A, Longitudinal sections of optic nerves derived brane lysates from both wild-type and mutant animals were from 45-d-old wild-type (WT) or CGT (CGTϪ/Ϫ) mice were labeled with subjected to immunoprecipitation with anti-Kv1.2 or Kv2.1 anti- the different combination of antibodies as indicated at the top of each ϩ bodies followed by immunoblotting with an antibody to Caspr2. panel. Caspr and K channels (Kv1.2) are labeled in green, and Caspr2 ϩ As depicted in Figure 3B, Caspr2 could be coimmunoprecipitated and Na channels (NaCh) are labeled in red. The location of a node is ␮ ϩ from CGTϪ/Ϫ mutant brains with an antibody against the Kv1.2 marked by an arrow. Scale bar, 20 m. B, Association of Caspr2 with K channels. Solubilized brain membranes from wild-type (WT)orCGT but not with an antibody to the control Kv2.1 subunit. These knock-out animals (CGT) were subjected to immunoprecipitation with ϩ experiments demonstrate that although Caspr2 and Kv1.2 are speciﬁc antibodies to either the Kv1.2 or Kv2.1 subunits of K channels, found in abnormal positions along the axon in the CNS of or to Caspr2 as indicated (IP). Note that from one brain that was used for each set of immunoprecipitations, 40% of the lysate was used for Kv1.2 and Kv2.1 antibodies, whereas 20% was used for Caspr2. The complexes were separated on SDS-gel and immunoblotted with an antibody to Caspr2. The location of molecular weight markers is shown on the right in kilodaltons.

CGTϪ/Ϫ mice, they are still associated in a physical complex as they are in wild-type animals. Caspr2/K ؉ channel complex during development ϩ The ﬁndings that Caspr2 and K channels are colocalized together in wild-type as well as in two paranodal mutant mice raise the question of whether the association between these molecules is important for their correct distribution along the nerve during ϩ development. In the rat sciatic nerve, K channels are ﬁrst detected at the nodes and paranodes and later in the juxtaparanodal region (Vabnick et al., 1999). Analyses of the distribution of Caspr2 and Kv1.2 in sciatic nerves of wild-type mice showed that both proteins could be detected from postnatal day 8 onward ϩ Figure 2. Distribution of K channels and members of the Caspr family in the juxtaparanodal region and to a lesser extent in the para- ϩ in peripheral nerves of wild-type and paranodal mutant mice. Teased nodes (Fig. 4). At this age, Na channels were found at all nodes, Ϫ Ϫ sciatic nerve preparations from wild-type (WT) or CGT (CGT / )or with Caspr already occupying the entire paranode. Caspr2 and contactin null mice (ConϪ/Ϫ) were labeled with different combination of ϩ antibodies as indicated at the top of each panel. Caspr and Caspr2 K channel clustering was frequently observed in a ring-like ϩ ϩ labeling is shown in green, whereas that of Na (NaCh) and K channels shape in the paranodal/juxtaparanodal border, often in an asym- (Kv1.2) is shown in red. Scale bar, 10 ␮m. metric manner (Fig. 4 and data not shown). In contrast, in sciatic Poliak et al. • Caspr-Dependent Localization of Caspr2 J. Neurosci., October 1, 2001, 21(19):7568–7575 7571

Figure 6. Caspr and Caspr2/Kϩ complex occupy mutually excluding domains along the axon. Wild-type (A–C and G–I)orCGTϪ/Ϫ (D–F) nerves were labeled for Caspr (red; B, E, H,) and Caspr2 (red; A, D, G)as indicated. The merged images of each experiment are shown on the right panels.InCGTϪ/Ϫ nerve, high levels of Caspr2 are detected in regions that lack Caspr (D, E, double arrows). Similarly, in wild-type animals, a line of Caspr staining in the juxtaparanodes does not overlap with Caspr2 Figure 4. Distribution of Caspr proteins and ion channels during PNS (G–I, single arrow). Note that the weak labeling of the outer aspects of the development. Teased sciatic nerves from wild-type (WT) and CGT mu- Ϫ Ϫ myelin sheath (D) is not specific and does not reflect Caspr staining in tant mice (CGT / ) were prepared at different days during development Schwann cells. as indicated on the top. Double staining was performed using Caspr ϩ ( green) and Kv1.2 (red) or Caspr2 ( green) and Na channel (NaCh, red) ϩ antibodies as indicated on the left of each row. Insets in the first panel show relocation of Caspr2/K channels to the paranodes, enables us to labeling for either Caspr (bottom inset) or Kv1.2 (top inset). Note that at ϩ assess whether Caspr family members could coexist along the day 8, K channels are found in both paranodes and juxtaparanodal ϩ regions. In the CGT mice, Caspr2 and K channels remain in the same axonal domains. To this end, we immunolabeled sciatic paranodes at all ages. Scale bar, 10 ␮m. nerves for Caspr and Caspr2 or Kv1.2 and analyzed those sites in which some paranodal accumulation of Caspr was still detected. In contrast to wild-type nerves that show a clear and sharp border between the paranodes and the juxtaparanodal region (Fig. 6C), in CGTϪ/Ϫ mutant nerves we observed a gradient of Caspr2 expression, with the highest concentration proximal to the node (Fig. 6D–F). At these sites, Caspr accumulated but did not overlap with the area that displayed the highest level of Caspr2 ϩ expression. This pattern of Caspr2 or K channel distribution was detected in all of the sites that also showed partial paranodal accumulation of Caspr in the adult nerve. In addition to its localization at the paranodes, in wild-type nerves Caspr was also ϩ found as a thin line bordered by parallel lines of K channels and Caspr2 along the internodes and just below the Schmidt–Lanter-

ϩ man incisures (Fig. 6) (Arroyo et al., 1999, 2001). Passing through Figure 5. Caspr2 and K channels appear together in axonal domains the juxtaparanodal region, this line of Caspr staining appears as during development. Merged images of double staining for Caspr2 ϩ ( green) and Kv1.2 (red) performed on teased sciatic nerves from wild-type a groove that was devoid of Caspr2 or K channels (Fig. 6G–L), (WT) and CGT mutant mice (CGTϪ/Ϫ) at different days during devel- demonstrating that even when present at the same domain, Caspr opment as indicated on the top.Theleft top panel shows the co- ϩ ␮ and the Caspr2/K complex are still separated. appearance of Caspr2 and Kv1.2 in heminodes. Scale bar, 10 m. In axons of adult wild-type mice, rings of Caspr were detected repeatedly along the internodes and at the ends of the juxtapara- ϩ nerves from CGTϪ/Ϫ mice, Caspr2 and K channels were de- nodal region (Fig. 7). Labeling for either Caspr2 or Kv1.2 re- tected in the paranodes at all ages examined. Double- vealed that these proteins accumulated at one side of the ring. immunofluorescence labeling for Caspr2 and Kv1.2 showed that Furthermore, the directionality of this unilateral accumulation these proteins appear together during development of the wild- seemed to be determined by the proximity to the node: Caspr and ϩ typeaswellasCGTϪ/Ϫ nerves (Fig. 5). These results demon- the Caspr2/K channel complex were always accumulated, keep- ϩ strate that Caspr2 and K channels emerge as a complex during ing the same polarity as that observed in the closest paranode/ the development of the nerve at the paranode/juxtaparanodal juxtaparanodal region. Altogether, these results show that Caspr border, which then moves from the paranodes to the juxtaparan- family members occupy mutually excluding domains along my- odal region. In CGTϪ/Ϫ fibers, where Caspr and contactin are elinated axons and suggest that the presence of Caspr along the not in place, the paranodal domain is not defined, and the axon at sites other than the paranodal junction may also serve as ϩ ϩ Caspr2/K channel complex remains localized adjacent to the a barrier for the movement of Caspr2 and K channels in the nodes. membrane. Caspr family members occupy mutually excluding Localization of protein 4.1B to the paranodal junction domains along the axons depends on members of the Caspr family The reduced paranodal accumulation of Caspr in sciatic nerves of The unique localization of members of the Caspr family along CGTϪ/Ϫ mice (Dupree et al., 1999; this work), along with the myelinated axons may require the attachment to the cytoskeleton 7572 J. Neurosci., October 1, 2001, 21(19):7568–7575 Poliak et al. • Caspr-Dependent Localization of Caspr2

ϩ Figure 7. Internodal localization of Caspr and Caspr2/K channel complex. Teased ﬁbers of adult mouse sciatic nerves were stained for Caspr (red; B, E, G, J, N) and Caspr2 ( green; A, D, H, K, M) or for Caspr ( green; P) and Kv1.2 (red; Q). Merged images for each set of labeling are also shown (C, F, I, L, O, R). Scale bar, 10 ␮m. Note that both Caspr2 and Kv1.2 are restricted to only one side of the Caspr ring, keeping the same polarity as that of the closest paranode.

Figure 8. Distribution of protein 4.1B in peripheral nerves of wild-type and paranodal mutant mice. Teased sciatic fibers from day 12 (A, B) and adult (C, D) wild-type animals, or 12-d-old (E, F) and adult (G, H) CGTϪ/Ϫ mice were double labeled with an antibody to protein 4.1B ( green) and an ϩ antibody to Na channel, Kv1.2, or Caspr (all in red) as indicated. Note that in adult CGTϪ/Ϫ nerve, paranodal concentration of protein 4.1B depends on the abnormal accumulation of the Caspr2/Kv1.2 complex at these sites. by binding to members of the protein 4.1 family. Such proteins juxtaparanodes. Protein 4.1B was observed at the paranodal junc- were shown to bind to the cytoplasmic domain of Caspr (Mene- tion and was occasionally detected also at the juxtaparanodal goz et al., 1997), as well as to neurexin IV, the Drosophila region (Fig. 8). The presence of protein 4.1B at the juxtapara- homolog of Caspr2 (Baumgartner et al., 1996). This family con- nodes increased with age and was more prominent in adult sists of four distinct proteins (Huang et al., 1993; Walensky et al., nerves, where it was localized with Caspr2 and Kv1.2. In contrast, 1998, 1999; Yamakawa et al., 1999; Parra et al., 2000), of which in sciatic nerve of 12-d-old CGT mice, protein 4.1B was distrib- only 4.1B was found at the paranodes in the CNS (Ohara et al., uted along the axon and weakly detected at the paranodes. When 2000) (L. Gollan and E. Peles, unpublished observation). To protein 4.1B was detected at the paranodes, it was found at sites determine the localization of protein 4.1B in the PNS, we first in which either Caspr or the Caspr2/Kv1.2 complexes were con- analyzed its distribution in sciatic nerve of wild-type animals at centrated. This was more conspicuous in adult sciatic nerves. postnatal day 12. At this developmental stage, Caspr was found at While in wild-type animals protein 4.1B was found at the paran- all paranodes, whereas Caspr2 was found only in part of the odes and the juxtaparanodal region, in CGTϪ/Ϫ nerves it was Poliak et al. • Caspr-Dependent Localization of Caspr2 J. Neurosci., October 1, 2001, 21(19):7568–7575 7573 diffused along the axon or detected at the paranodes together (Arroyo et al., 1999; Pedraza et al., 2001). This, in turn, may allow ϩ with the Caspr2/Kv1.2 complex (Fig. 8). In summary, these re- the K channels to accumulate as stripes between the Caspr rings sults indicate that mislocalization of both Caspr proteins in at the paranodes. This notion is well supported by ultrastructural galactolipids-deficient mice is also accompanied by aberrant dis- data, revealing the presence of distinct rows of particles at the tribution of protein 4.1B, suggesting that members of the Caspr paranodal axolemma that are regularly positioned between the family may regulate the localization of 4.1B in myelinated axons. paranodal loops (Rosenbluth, 1995). In addition to its paranodal localization, the Caspr/contactin complex is also present as a thin DISCUSSION line at the inner mesaxon and below the Schmidt–Lanterman Sequential appearance of Caspr and Caspr2 incisures (Menegoz et al., 1997; Arroyo et al., 1999; Rios et al., during development 2000). When this line was present at the juxtaparanodes, it was constantly detected in a groove, which was devoid of Caspr2 or We have shown previously that Caspr2 is localized and associates ϩ ϩ Kv1.2. The observation that Caspr and the Caspr2/K channel with members of the Shaker K channels family, Kv1.1, Kv1.2, complex always occupy excluding zones in the PNS is consistent and their Kv␤2 subunit at the juxtaparanodal region (Poliak et al., with recent data demonstrating that although in some spinal cord 1999). This association occurs indirectly and probably involves ϩ axons K channels were found at the paranodal region, they did the interaction of Caspr2 with a PDZ-containing protein (Poliak not overlap with Caspr labeling (Rasband and Trimmer, 2001). et al., 1999). In this work, we compared the distribution of Caspr2 and Kv1.2 during development of peripheral nerves of wild-type Localization of Caspr2 in myelinated axons depends and mutant mice with paranodal abnormalities. At all ages tested, ϩ on Caspr-containing domains Caspr2 always colocalized with K channels. In sciatic nerves of In addition to their proposed role in anchoring the myelin to the wild-type animals, Caspr2 could be detected as early as day 8 at axon and in forming a partial barrier that isolates the periaxonal the paranodes and the juxtaparanodal region, where it had al- space from the nodes, the paranodal junctions were proposed to ready been colocalized with Kv1.2. Comparatively, at this age, physically demarcate boundaries that limit the lateral diffusion of Caspr was present at all paranodes. As development progressed, membrane components (Rosenbluth, 1995; Peles and Salzer, Caspr2 translocated to the adjacent juxtaparanodal region until it 2000). The mislocalization of Kv1.1 (Dupree et al., 1999), Caspr2, became completely absent from the paranodes in the adult. This Kv1.2, and protein 4.1B in CGTϪ/Ϫ mice described here strongly pattern is in agreement with a previous study, demonstrating a ϩ supports such barrier-like function. Surprisingly, in addition to role for these paranodal K channels in preventing repetitive ϩ the paranodes, we observed an accumulation of Caspr2 and K firing during development of peripheral nerves (Vabnick et al., channels near rings of Caspr at the end of the juxtaparanodes and 1999). In contrast to the wild-type nerves, in the galactolipids- along the internodal region, suggesting the existence of such deficient mice, Caspr2 and Kv1.2 were initially detected at the structures also at these domains. In principal, the accumulation of paranodes but did not relocate to the juxtaparanodal region and ϩ Caspr2/K channels at these sites could reflect a stopping of remained mostly paranodal in the adult (94 and 96% of the sites, moving complexes at the axonal membrane, or, less likely, an respectively). A similar distribution was reported previously for ϩ anchorage and sorting site for newly arriving Caspr2/K channel Kv1.1 in the adult sciatic nerve (85% of the sites examined) complexes. We propose that these rings are similar to those that (Dupree et al., 1999). Taken together, these results demonstrate ϩ are squeezed together in the paranodes and most likely represent that the entire complex, which contains Caspr2 and distinct K similar barrier properties as found at the paranodes to restrict the channel subunits, is abnormally found at the paranodes in this movement of axonal components, including cell adhesion recep- mutant. Furthermore, although in the CNS, Caspr2 and Kv1.2 tors and ion channels. Remarkably, the accumulation of were less confined to the juxtaparanodal region; they were invari- ϩ Caspr2/K channels near such sites occurred in only one side of ably colocalized to ectopic sites and still generated a stable the ring and was always directed toward the closest node, indi- complex. The transient appearance of Caspr2 and Kv1.2 in the cating that the internodal region itself is polarized. paranodes during development of peripheral nerves raises the possibility that these proteins may be targeted directly to this site. Molecular mechanisms of paranodal phenotype of In this regard, the paranodal junction may serve as a sorting galactolipids-deficient mice center much like the proposed role of tight junctions in epithelial To date, three transmembrane proteins were found to reside at cells (Zimmermann, 1996; Zahraoui et al., 2000). the paranodal junction. These include Caspr (Einheber et al., Caspr and Caspr2 occupy mutually excluding domains 1997; Menegoz et al., 1997) and the two immunoglobulin cell in myelinated axons adhesion molecules, contactin and NF155 (Rios et al., 2000; Tait ϩ During development of wild-type peripheral nerves, Caspr2/K et al., 2000). Caspr and contactin form a complex (Peles et al., channel complexes are excluded from the paranodes, which are 1997) that is found at the paranodes (Rios et al., 2000), and both already occupied by Caspr. This observation is further supported are essential for the generation of the paranodal septate-like by analyses of sciatic nerves of CGTϪ/Ϫ mice. In contrast to junction (Bhat et al., 2001; Boyle et al., 2001). Similarly, wild-type nerves, which exhibit a sharp boundary between Caspr2 galactolipids-deficient mice also exhibit paranodal abnormalities and Caspr zones, no such border was detected at sites that still and lack the transverse bands and Caspr in their paranodes showed paranodal accumulation of Caspr in CGTϪ/Ϫ mutants (Dupree et al., 1998, 1999). Here, we extended these observations (Fig. 6). At these sites, Caspr distinctly occupied only the distal and show that the paranodes of peripheral myelinated nerves of part of the paranodal region (i.e., in respect to the nodes) and CGTϪ/Ϫ mice are also lacking contactin and NF155. This result avoided the more proximal part that contained a high concentra- demonstrates that disruption of normal axo-glial contact in these tion of Caspr2. It should be noted that paranodal accumulation of mice is accompanied by the disappearance of all known compo- Caspr is composed of a number of rings that represent each turn nents of the paranodal junction (Fig. 9). Furthermore, disruption of the myelin wrap and thus does not represent a uniform domain of axon–glia contact in CGT mice not only results in mislocal- 7574 J. Neurosci., October 1, 2001, 21(19):7568–7575 Poliak et al. • Caspr-Dependent Localization of Caspr2

invariably appear together during development of peripheral nerves. The correct localization of this complex in myelinated axons depends on the formation of Caspr-containing structural barriers along the axon, whereas the disruption of these barriers in CGT and contactin null mice results in its mislocalization. Future analyses of Caspr2 null mice will reveal whether it is ϩ indeed essential for the proper localization of K channels to the juxtaparanodal region.

Figure 9. Molecular composition of the nodal region in peripheral nerves. Schematic summary illustrating the distribution of cell adhesion REFERENCES molecules, ion channels, and cytoplasmic proteins along myelinated Ϫ Ϫ Arroyo EJ, Xu YT, Zhou L, Messing A, Peles E, Chiu SY, Scherer SS nerves of wild-type and galactolipids-deficient mice (CGT / ). The (1999) Myelinating Schwann cells determine the internodal localiza- nodal region is divided into several distinct domains, including the node tion of Kv1.1, Kv1.2, Kvbeta2, and Caspr. J Neurocytol 28:333–347. of Ranvier, paranodal (PN), juxtaparanodal (JXT), and internodal re- Arroyo EJ, Xu YT, Poliak S, Watson M, Peles E, Scherer SS (2001) ϩ gions. In both wild-type and CGTϪ/Ϫ mutant nerves, Na channels are Internodal specializations of myelinated axons in the CNS. Tissue Cell present at the nodes of Ranvier. In wild-type animals, Caspr and contac- Res 305:53–66. tin label the axonal side of the paranodal junction, whereas neurofascin Baumgartner, S, Littleton JT, Broadie, K, Bhat MA, Harbecke, R, 155 (NF155) is found at the glial loops. Caspr2 and Kv1.2 are present as Lengyel JA, Chiquet-Ehrismann, R, Prokop, A, Bellen HJ (1996) A Drosophila neurexin is required for septate junction and blood-nerve a complex at the juxtaparanodal area, which is located under the compact barrier formation and function. Cell 87:1059–1068. myelin. In contrast to the wild-type nerves, in peripheral nerves of the Bekele-Arcuri Z, Matos MF, Manganas L, Strassle BW, Monaghan MM, CGTϪ/Ϫ mutant, Caspr, contactin, and neurofascin are no longer present Rhodes KJ, Trimmer JS (1996) Generationϩ and characterization of at the paranodal junction. Caspr is primarily distributed along the axon, subtype-specific monoclonal antibodies to K channel alpha- and beta- whereas NF155 is weakly diffused at the paranodal loops area. Disruption subunit polypeptides. Neuropharmacology 35:851–865. of the axo-glial junction results in relocation of Caspr2 and Kv1.2 from the Berglund EO, Murai KK, Fredette B, Sekerkova G, Marturano B, Weber juxtaparanodal region to the adjacent paranode. In wild-type nerves, the L, Mugnaini E, Ranscht B (1999) Ataxia and abnormal cerebellar cytoplasmic protein 4.1B is present at both the paranodal and the jux- microorganization in mice with ablated contactin gene expression. Neuron 24:739–750. taparanodal regions. In CGTϪ/Ϫ fibers, protein 4.1B is either diffused Bhat AM, Rios JC, Lu Y, Garcia-Fresco GP, Ching W, St. Martin M, Li along the axon or found at the paranodes along with Caspr2. This J, Einheber S, Chesler M, Rosenbluth J, Salzer JL, Bellen HJ (2001) summary is based on the present work, as well as on a previous work by Axon-glia interactions and domain organization of myelinated axons Popko and colleagues (Dupree et al., 1999). requires Neurexin IV/Caspr/Paranodin. Neuron 30:369–383. Bosio A, Bussow H, Adam J, Stoffel W (1998) Galactosphingolipids and axono-glial interaction in myelin of the central nervous system. Cell ization of membrane junctional components but also affects the Tissue Res 292:199–210. distribution of downstream cytoskeletal elements such as protein Boyle MET, Berglund EO, Murai KK, Weber L, Peles E, Ranscht B (2001) Contactin orchestrates assembly of the septate-like junctions at 4.1. In the wild-type nerve, protein 4.1B was first detected with the paranode in myelinated peripheral nerve. Neuron 30:385–397. Caspr at the paranodes and only at a later stage with Caspr2 at Coetzee T, Fujita N, Dupree J, Shi R, Blight A, Suzuki K, Popko B the juxtaparanodal region. In contrast, in the absence of Caspr in (1996) Myelination in the absence of galactocerebroside and sulfatide: Ϫ Ϫ normal structure with abnormal function and regional instability. Cell CGT / mice, protein 4.1B follows the distribution of Caspr2 86:209–219. and remains at the paranodes. These results suggest that protein Davis JQ, Lambert S, Bennett V (1996) Molecular composition of the 4.1B may be recruited to sites that contain Caspr and Caspr2, node of Ranvier: identification of ankyrin binding cell adhesion molecules neurofascin (mucinϩ/third FNIII domain-) and NrCAM at nodal where it may assist in maintaining functional domains by linking axon segments. J Cell Biol 135:1355–1367. these proteins to the axonal cytoskeleton. Dupree JL, Coetzee T, Blight A, Suzuki K, Popko B (1998) Myelin galactolipids are essential for proper node of Ranvier formation in the Several mechanisms have been suggested previously to account CNS. J Neurosci 18:1642–1649. for the mislocalization of Caspr in CGTϪ/Ϫ nerves, including the Dupree JL, Girault JA, Popko B (1999) Axo-glial interactions regulate direct binding of Caspr to GalC and/or sulfatide, as well as the localization of axonal paranodal proteins. J Cell Biol 147:1145–1152. misrouting of a yet unidentified glial ligand to the glial loops Einheber S, Zanazzi G, Ching W, Scherer S, Milner TA, Peles E, Salzer (Dupree et al., 1999; Popko, 2000). The latter is based on the JL (1997) The axonal membrane protein Caspr, a homologue of neur- proposed role of galactolipids in the formation of lipid rafts and exin IV, is a component of the septate-like paranodal junctions that assemble during myelination. J Cell Biol 139:1495–1506. the organization of the myelin membranes (Kramer et al., 1997; Huang JP, Tang CJ, Kou GH, Marchesi VT, Benz EJ, Tang TK (1993) Kim and Pfeiffer, 1999; Simons et al., 2000). The finding that Genomic structure of the locus encoding protein 4.1. Structural basis NF155 is reduced or absent from the paranodal loops of the for complex combinational patterns of tissue-specific alternative RNA splicing. J Biol Chem 268:3758–3766. galactolipids-deficient mice reported here raises the possibility Kim T, Pfeiffer SE (1999) Myelin glycosphingolipid/cholesterol- that localization of the Caspr/contactin complex at the paranodal enriched microdomains selectively sequester the non-compact myelin proteins CNP and MOG. J Neurocytol 28:281–293. junction may depend on the presence of NF155 at this site. Kramer EM, Koch T, Niehaus A, Trotter J (1997) Oligodendrocytes According to this model, it is likely that during development, the direct glycosyl phosphatidylinositol-anchored proteins to the myelin Caspr/contactin complex is recruited to axo-glial contact sites at sheath in glycosphingolipid-rich complexes. J Biol Chem 272:8937–8945. the mesaxon and the paranodal loops by NF155. Nevertheless, Menegoz M, Gaspar P, Le Bert M, Galvez T, Burgaya F, Palfrey C, Ezan the observation that NF155 is reduced in or absent from the P, Arnos F, Girault JA (1997) Paranodin, a glycoprotein of neuronal paranodes of contactin null mice (Boyle et al., 2001) demonstrates paranodal membranes. Neuron 19:319–331. Ohara R, Yamakawa H, Nakayama M, Ohara O (2000) Type II brain 4.1 that proper localization of all putative adhesion components of (4.1B/KIAA0987), a member of the protein 4.1 family, is localized to the paranodal junction is interdependent, rendering still feasible neuronal paranodes. Brain Res Mol Brain Res 85:41–52. Parra M, Gascard P, Walensky LD, Gimm JA, Blackshaw S, Chan N, the possibility that other mechanisms may account for the para- Takakuwa Y, Berger T, Lee G, Chasis JA, Snyder SH, Mohandas N, nodal defect observed in the galactolipids-deficient mice. Conboy JG (2000) Molecular and functional characterization of pro- In summary, we have extended previous observations which tein 4.1B, a novel member of the protein 4.1 family with high level, focal ϩ expression in brain. J Biol Chem 275:3247–3255. demonstrated that Caspr2 forms a complex with K channels at Pedraza L, Huang JK, Colman DR (2001) Organizing principles of the the juxtaparanodal region and have found that these molecules axoglial apparatus. Neuron 30:335–344. Poliak et al. • Caspr-Dependent Localization of Caspr2 J. Neurosci., October 1, 2001, 21(19):7568–7575 7575

Peles E, Salzer JL (2000) Functional domains in myelinated axons. Curr Tait S, Gunn-Moore F, Collinson JM, Huang J, Lubetzki C, Pedraza L, Opin Neurobiol 10:558–565. Sherman DL, Colman DR, Brophy PJ (2000) An oligodendrocyte cell Peles E, Nativ M, Lustig M, Grumet M, Schilling J, Martinez R, Plowman adhesion molecule at the site of assembly of the paranodal axo-glial GD, Schlessinger J (1997) Identification of a novel contactin- junction. J Cell Biol 150:657–666. associated transmembrane receptor with multiple domains implicated Vabnick I, Shrager P (1998) Ion channel redistribution and function in protein-protein interactions. EMBO J 16:978–988. during development of the myelinated axon. J Neurobiol 37:80–96. Poliak S, Gollan L, Martinez R, Custer A, Einheber S, Salzer JL, Vabnick I, Trimmer JS, Schwarz TL, Levinson SR, Risal D, Shrager P Trimmer JS, Shrager P, Peles E (1999) Caspr2, a new member of the (1999) Dynamic potassium channel distributions during axonal devel- neurexin superfamily, is localized at the juxtaparanodes of myelinated opment prevent aberrant firing patterns. J Neurosci 19:747–758. axons and associates with Kϩ channels. Neuron 24:1037–1047. Walensky LD, Gascard P, Fields ME, Blackshaw S, Conboy JG, Mohan- Popko B (2000) Myelin galactolipids: mediators of axo-glial interac- das N, Snyder SH (1998) The 13-kD FK506 binding protein, FKBP13, tions? Glia 29:149–153. interacts with a novel homologue of the erythrocyte membrane cy- Rasband MN, Shrager P (2000) Ion channel sequestration in central toskeletal protein 4.1. J Cell Biol 141:143–153. nervous system axons. J Physiol (Lond) 525:63–73. Walensky LD, Blackshaw S, Liao D, Watkins CC, Weier HU, Parra M, Rasband MN, Trimmer JS (2001) Subunit composition and novel local- Huganir RL, Conboy JG, Mohandas N, Snyder SH (1999) A novel ization of Kϩ channels in spinal cord. J Comp Neurol 429:166–176. neuron-enriched homolog of the erythrocyte membrane cytoskeletal Rasband MN, Peles E, Trimmer JS, Levinson SR, Lux SE, Shrager P protein 4.1. J Neurosci 19:6457–6467. (1999) Dependence of nodal sodium channel clustering on paranodal Wang H, Kunkel DD, Martin TM, Schwartzkroin PA, Tempel BL (1993) axoglial contact in the developing CNS. J Neurosci 19:7516–7528. Heteromultimeric Kϩ channels in terminal and juxtaparanodal regions Rhodes KJ, Strassle BW, Monaghan MM, Bekele-Arcuri Z, Matos MF, of neurons. Nature 365:75–79. ␤ Trimmer JS (1997) Association and colocalization of the Kvϩ 1 and Waxman SG, Ritchie JM (1993) Molecular dissection of the myelinated Kv␤2 ␤-subunits with Kv1 ␣-subunits in mammalian brain K channel axon. Ann Neurol 33:121–136. complexes. J Neurosci 17:8246–8258. Winckler B, Forscher P, Mellman I (1999) A diffusion barrier maintains Rios JC, Melendez-Vasquez CV, Einheber S, Lustig M, Grumet M, distribution of membrane proteins in polarized neurons. Nature Hemperly J, Peles E, Salzer JL (2000) Contactin-associated protein 397:698–701. (Caspr) and contactin form a complex that is targeted to the paranodal Yamakawa H, Ohara R, Nakajima D, Nakayama M, Ohara O (1999) junctions during myelination. J Neurosci 20:8354–8364. Molecular characterization of a new member of the protein 4.1 family Rosenbluth J (1995) Glial membranes and axoglial junctions. In: Neu- (brain 4.1) in rat brain. Brain Res Mol Brain Res 70:197–209. roglia (Kettenmann H, Ransom BR, eds), pp 613– 633. New York: Zahraoui A, Louvard D, Galli T (2000) Tight junction, a platform for Oxford UP. trafficking and signaling protein complexes. J Cell Biol 151:31–36. Simons M, Kramer EM, Thiele C, Stoffel W, Trotter J (2000) Assembly Zimmermann H (1996) Accumulation of synaptic vesicle proteins and of myelin by association of proteolipid protein with cholesterol- and cytoskeletal specializations at the peripheral node of Ranvier. Microsc galactosylceramide-rich membrane domains. J Cell Biol 151:143–154. Res Tech 34:462–473.